Predicting Remaining Useful Life for a Computer System Using a Stress-Based Prediction Technique

1. A method for predicting a remaining useful life (RUL) for a component in a set of components within a computer system, wherein the set of components includes a monitored component, the method comprising, in at least one computer, performing operations for: collecting values of at least one degradation-related parameter associated with the operation of the monitored component within the computer system, wherein the degradation-related parameter is a direct measurement of a degree of degradation of the monitored component; collecting values of at least one stress-based parameter from the computer system, wherein the stress-based parameter measures an accumulative stress in the operating environment of the set of components which can cause degradation of the set of components; and predicting an RUL for a component in the set of components using at least a predicted first RUL for the monitored component and a predicted second RUL for the set of components, wherein the first RUL is predicted based at least on the values of the at least one degradation-related parameter and the second RUL is predicted based at least on the values of the at least one stress-based parameter; wherein predicting the RUL for the component in the set of components using at least the first RUL and the second RUL involves: determining if the first RUL indicates that the monitored component is not close to failure and if the second RUL indicates that the set of components is close to failure; and if so, predicting the RUL for the monitored component based on the first RUL when the first RUL is associated with a first confidence level; and otherwise, predicting the RUL for the monitored component based on the second RUL when the first RUL is associated with a second confidence level that is lower than the first confidence level.

2. The method of claim 1 , further comprising: predicting the first RUL for the monitored component based at least on the values of the at least one degradation-related parameter; and predicting the second RUL for the set of components based at least on the values of the at least one stress-based parameter.

3. The method of claim 1 , wherein predicting an RUL for a component in the set of components based at least on the first RUL and the second RUL involves: determining if the first RUL indicates that the monitored component is close to failure and if the second RUL indicates that the set of components is not close to failure; and if so, predicting an RUL for the monitored component based on the first RUL, wherein the predicted RUL for the monitored component is shorter than the second RUL.

4. The method of claim 3 , wherein the method further comprises replacing the monitored component proactively.

5. The method of claim 1 , wherein the method further comprises predicting an RUL for the set of components other than the monitored component based on the second RUL, wherein the predicted RUL for the set of components is shorter than the first RUL.

6. The method of claim 1 , wherein predicting an RUL for a component in the set of components based at least on the first RUL and the second RUL involves: determining if the first RUL indicates that the monitored component is close to failure sooner than expected and if the second RUL indicates that the set of components is close to failure; and if so, determining that the set of components is approaching failure sooner than expected as a result of the operating environment stress which is more severe than expected.

7. The method of claim 1 , wherein the method further comprises validating signals from sensors which are used to collect the values of at least one degradation-related parameter and values of at least one stress-based parameter.

8. The method of claim 1 , wherein predicting the RUL can take place at least one of the following locations: on the computer system; on a loghost machine external to the computer system; and within a centralized service center.

9. The method of claim 1 , wherein the stress-based parameter includes one or more of the following: temperature; voltage; current; vibration; and humidity.

10. The method of claim 6 , wherein the method further comprises proactively replacing the set of components.

11. The method of claim 7 , wherein validating the signals from the sensors involves using a multivariate state estimation technique (MSET).

12. An apparatus that predicts a remaining useful life (RUL) for a component in a set of components within a computer system, wherein the set of components includes a monitored component, comprising: a collection mechanism coupled to the computer system, wherein the collection mechanism is configured to collect values of at least one degradation-related parameter associated with the operation of the monitored component within the computer system, wherein the degradation-related parameter is a direct measurement of a degree of degradation of the monitored component; wherein the collection mechanism is further configured to collect values of at least one stress-based parameter from the computer system, wherein the stress-based parameter measures an accumulative stress in the operating environment of the set of components which can cause degradation of the set of components; and a first RUL prediction mechanism coupled to the collection mechanism, wherein the first RUL prediction mechanism is configured to predict an RUL for a component in the set of components using at least a predicted first RUL for the monitored component and a predicted second RUL for the set of components, wherein the first RUL is predicted based at least on the values of the at least one degradation-related parameter and the second RUL is predicted based at least on the values of the at least one stress-based parameter; wherein the first RUL prediction mechanism is configured to: determine if the first RUL indicates that the monitored component is not close to failure and if the second RUL indicates that the set of components is close to failure; and if so, predict the RUL for the monitored component based on the first RUL when the first RUL is associated with a first confidence level; and otherwise, predict the RUL for the monitored component based on the second RUL when the first RUL is associated with a second confidence level that is lower than the first confidence level.

13. The apparatus of claim 12 , further comprising: a second RUL prediction mechanism configured to predict the first RUL for the monitored component based at least on the values of the at least one degradation-related parameter; and a third RUL prediction mechanism configured to predict the second RUL for the set of components based at least on the values of the at least one stress-based parameter.

14. The apparatus of claim 12 , wherein the first RUL prediction mechanism is further configured to: determine if the first RUL indicates that the monitored component is close to failure and if the second RUL indicates that the set of components is not close to failure; and if so, predict the RUL for the monitored component based on the first RUL, wherein the predicted RUL for the monitored component is shorter than the second RUL.

15. The apparatus of claim 12 , wherein the first RUL prediction mechanism is further configured to: determine if the first RUL indicates that the monitored component is close to failure sooner than expected and if the second RUL indicates that the set of components is close to failure; and if so, determine that the set of components is approaching failure sooner than expected as a result of the operating environment stress which is more severe than expected.

16. The apparatus of claim 12 , wherein the first RUL prediction mechanism can reside on at least one of the following locations: the computer system; a loghost machine external to the computer system; and a centralized service center.

17. A computer-readable storage medium storing instructions that when executed by a computer cause the computer to perform a method for predicting a remaining useful life (RUL) for a component in a set of components within a computer system, wherein the set of components includes a monitored component, the method comprising: collecting values of at least one degradation-related parameter associated with the operation of the monitored component within the computer system, wherein the degradation-related parameter is a direct measurement of a degree of degradation of the monitored component; collecting values of at least one stress-based parameter from the computer system, wherein the stress-based parameter measures an accumulative stress in the operating environment of the set of components which can cause degradation of the set of components; and predicting an RUL for a component in the set of components using at least a predicted first RUL for the monitored component and a predicted second RUL for the set of components, wherein the first RUL is predicted based at least on the values of the at least one degradation-related parameter and the second RUL is predicted based at least on the values of the at least one stress-based parameter; wherein predicting the RUL for the component in the set of components using at least the first RUL and the second RUL involves: determining if the first RUL indicates that the monitored component is not close to failure and if the second RUL indicates that the set of components is close to failure; and if so, predicting the RUL for the monitored component based on the first RUL when the first RUL is associated with a first confidence level; and otherwise, predicting the RUL for the monitored component based on the second RUL when the first RUL is associated with a second confidence level that is lower than the first confidence level.

18. The computer-readable storage medium of claim 17 , wherein predicting the RUL for a component in the set of components involves: predicting the first RUL for the monitored component based at least on the values of the at least one degradation-related parameter; and predicting the second RUL for the set of components based at least on the values of the at least one stress-based parameter.